The Method for Steel Surface Defect Detection and Classification Based on the Improved YOLOv7

To effectively improve the accuracy and efficiency of detecting defects on steel surfaces, this paper proposes an improved YOLOv7-based steel surface defect detection algorithm. First, by introducing the Bidirectional Feature Pyramid Network (BiFPN), more effective information fusion is achieved between features at different scales, significantly enhancing the model's ability to extract multi-scale features. Second, to improve the accuracy of object localization, the paper adopts a global K-Means algorithm to optimize the anchor box generation strategy, enabling the generation of precise and appropriate anchor boxes based on the distribution characteristics of steel surface defects. Additionally, the paper introduces the Efficient Channel Attention Network (ECANet), which further enhances the model's ability to focus on key regions, helping the network focus on more discriminative feature areas, thereby improving classification accuracy and detection precision. Experimental results show that the improved model achieves an average precision(mAP) of 77.9 % on the public NEU-DET dataset, representing a 5 % improvement over the original YOLOv7 model. This method demonstrates excellent performance in complex industrial environments, providing reliable technical support for quality control in steel production lines.